Artificial Immune Systems: A New Computational Intelligence Approach
Published byModified over 5 years ago
Presentation on theme: "Artificial Immune Systems: A New Computational Intelligence Approach"— Presentation transcript:
1 Artificial Immune Systems: A New Computational Intelligence Approach New Trends in Intelligent Information Processingand Web Mining.Zakopane, Poland, June 2-5, 2003Jonathan TimmisComputing LaboratoryUniversity of KentCT2 7NF. UK.
2 Novel paradigms are proposed and accepted not necessarily for being faithfulto their sources of inspiration, but forbeing useful and feasible
3 What do I want to achieve? Give you a taster of what AIS is all aboutDefine an AISWhy do we find the immune system useful?Explain what AIS areShow you where they are being usedSome high level case studiesComments for the futureI won’tTalk about all areas of AIS and applicationsTalk too much about how AIS relate to other bioinspired ideas (although I will mention it)Go into too much detail: this is an introduction
4 Outline What are AIS? Useful immunology Thinking about AIS Application Areas and Case StudiesThe Future
5 Why the Immune System? Recognition Robustness Feature extraction Anomaly detectionNoise toleranceRobustnessFeature extractionDiversityReinforcement learningMemory; Dynamically changing coverageDistributedMulti-layeredAdaptive· Uniqueness: each individual possesses its own immune system, with its particular vulnerabilities and capabilities;· Diversity: there is a large amount of types of elements (cells, molecules, proteins, etc.) that altogether perform the same role of protecting the body from malefic invaders. Additionally, there are different fronts of defense, like innate and adaptive immunity;· Disposability (robustness): no single component of the natural immune system is essential for its functioning. Cell death is usually balanced by cell production;· Autonomy: the immune system does not require outside management or maintenance. It autonomously classifies and eliminates pathogens, and it repairs itself by replacing damaged cells;· Multilayered: multiple layers of different mechanisms are combined to provide high overall security, as summarized in Figure 2.5 (Section 2.3);· No secure layer: any cell of the human body can be attacked by the immune system, including those of the immune system itself;· Recognition of foreigners: the (harmful) molecules that are not native to the body are recognized and eliminated by the immune system;· Anomaly detection: the immune system can detect and react to pathogens that the body has never encountered before;· Dynamically changing coverage: as the immune system can not maintain a set of cells and molecules large enough to detect all pathogens, it makes a trade-off between space and time. It maintains a circulating pool of lymphocytes that is constantly being changed through cell death, production and reproduction;· Distributability: the immune cells, molecules and organs are distributed all over the body and, most importantly, are not subject to any centralized control;· Imperfect detection (noise tolerance): an absolute recognition of the pathogens is not required, hence the system is flexible;· Reinforcement learning and memory: the immune system can “learn” the structures of pathogens. It retains the ability to recognize previously seen pathogens through immune memory, so that future responses to the same pathogens are faster and stronger; and· An arms race: the vertebrate immune system replicates cells to deal with replicating pathogens, otherwise the pathogens would quickly overwhelm the immune defenses.
6 A DefinitionAIS are adaptive systems inspired by theoretical immunology and observed immune functions, principles and models, which are applied to complex problem domains
7 Some HistoryDeveloped from the field of theoretical immunology in the mid 1980’s.Suggested we ‘might look’ at the IS1990 – Bersini first use of immune algos to solve problemsForrest et al – Computer Security mid 1990’sHunt et al, mid 1990’s – Machine learning
8 Scope of AIS:Computer Security(Forrest’94’96’98, Kephart’94, Lamont’98’01,02, Dasgupta’99’01, Bentley’00’01,02)Anomaly Detection (Dasgupta’96’01’02)Fault Diagnosis (Ishida’92’93, Ishiguro’94)Data Mining & Retrieval (Hunt’95’96, Timmis’99’01, ’02)Pattern Recognition (Forrest’93, Gibert’94, de Castro ’02)Adaptive Control (Bersini’91)
9 Scope of AIS (Cont……): Job shop Scheduling (Hart’98, ’01, ’02) Chemical Pattern Recognition (Dasgupta’99)Robotics (Ishiguro’96’97,Singh’01)Optimization (DeCastro’99,Endo’98, de Castro ’02)Web Mining (Nasaroui’02)Fault Tolerance (Tyrrell, ’01, ’02, Timmis ’02)Autonomous Systems (Varela’92,Ishiguro’96)Engineering Design Optimization (Hajela’96 ’98, Nunes’00)And so on …
10 Outline What are AIS? Useful immunology Thinking about AIS Application Areas and Case StudiesThe Future
11 Role of the Immune System Protect our bodies from pathogen and virusesPrimary immune responseLaunch a response to invading pathogensSecondary immune responseRemember past encountersFaster response the second time around
13 Immune cells There are two primarily types of lymphocytes: B-lymphocytes (B cells)T-lymphocytes (T cells)Others types include macrophages, phagocytic cells, cytokines, etc.
14 Self/Non-Self Recognition Immune system needs to be able to differentiate between self and non-self cellsAntigenic encounters may result in cell death, thereforeSome kind of positive selectionSome element of negative selection
15 AntigenSubstances capable of starting a specific immune response commonly are referred to as antigensThis includes some pathogens such as viruses, bacteria, fungi etc .
16 Immune Pattern Recognition The immune recognition is based on the complementarity between the binding region of the receptor and a portion of the antigen called epitope.Antibodies present a single type of receptor, antigens might present several epitopes.This means that each antibody can recognize a single antigen
18 Main Properties of Clonal Selection (Burnet, 1978) Elimination of self antigensProliferation and differentiation on contact of mature lymphocytes with antigenRestriction of one pattern to one differentiated cell and retention of that pattern by clonal descendants;Generation of new random genetic changes, subsequently expressed as diverse antibody patterns by a form of accelerated somatic mutation
19 Immune Network Theory Idiotypic network (Jerne, 1974) B cells co-stimulate each otherTreat each other a bit like antigensCreates an immunological memoryMention Bersinis' principles
20 Reinforcement Learning and Immune Memory Repeated exposure to an antigen throughout a lifetimePrimary, secondary immune responsesRemembers encountersNo need to start from scratchMemory cellsContinuous learning
22 Immune System: Summary Define host (body cells) from external entities.When an entity is recognized as foreign (or dangerous)- activate several defense mechanisms leading to its destruction (or neutralization).Subsequent exposure to similar entity results in rapid immune response.Overall behavior of the immune system is an emergent property of many local interactions.So it is useful?
23 Outline What are AIS? Useful immunology Thinking about AIS Application Areas and Case StudiesThe Future
24 Artificial Immune Systems AIS are adaptive systems inspired by theoretical immunology and observed immune functions, principles and models, which are applied to complex problem domains
25 This SectionGeneral Framework for describing and constructing AIS modelsA short review of where AIS are used todayCan not cover them all, far too manyAlso we are not experts in all application areas !Where are AIS headed?
26 What do want from a Framework? In a computational world we work with representations and processes. Therefore, we need:To be able to describe immune system componentsBe able to describe their interactionsQuite high level abstractionsCapture general purpose processes that can be applied to various areas
27 General Framework for AIS Immune AlgorithmsAffinity MeasuresRepresentationApplication Domain
28 Representation – Shape Space Describe the general shape of a moleculeDescribe interactions between moleculesDegree of binding between molecules
29 Representation Vectors Ab = Ab1, Ab2, ..., AbL Ag = Ag1, Ag2, ..., AgLReal-valued shape-spaceInteger shape-spaceBinary shape-spaceSymbolic shape-space· Real-valued shape-space: the attribute strings are real-valued vectors;· Integer shape-space: the attribute strings are composed of integer values;· Hamming shape-space: composed of attribute strings built out of a finite alphabet of length k;· Symbolic shape-space: usually composed of different types of attribute strings where at least one of them is symbolic, such as a ‘name’, a ‘color’, etc.Assume the general case in which an antibody molecule is represented by the set of coordinates Ab = Ab1, Ab2, ..., AbL, and an antigen is given by Ag = Ag1, Ag2, ..., AgL, where boldface letters correspond to a string.
30 Define their Interaction Define the term AffinityDistance measures such as Hamming, Manhattan etc. etc.Affinity Threshold
32 Negative Selection (NS) Algorithms Forrest 1994: Idea taken from the negative selection of T-cells in the thymusApplied initially to computer securitySplit into two parts:CensoringMonitoring
33 Clonal Selection Algorithm (de Castro & von Zuben, 2001) 1. Initialisation: Randomly initialise a population (P)2. Antigenic Presentation: for each pattern in Ag, do:2.1 Antigenic binding: determine affinity to each P’2.2 Affinity maturation: select n highest affinity from P and clone and mutate prop. to affinity with Ag, then add new mutants to P3. Metadynamics:3.1 select highest affinity P to form part of M3.2 replace n number of random new ones4. Cycle: repeat 2 and 3 until stopping criteria1. Randomly initialize a population of individuals (P);2. For each pattern of S, present it to the population P and determine its affinity with each element of the population P;3. Select n1 highest affinity elements of P and generate copies of these individuals proportionally to their affinity with the antigen. The higher the affinity, the higher the number of copies, and vice-versa;4. Mutate all these copies with a rate proportional to their affinity with the input pattern: the higher the affinity, the smaller the mutation rate, and vice-versa;5. Add these mutated individuals to the population P and re-select n2 of these maturated (optimized) individuals to be kept as the memory M of the system;6. Replace a number n3 of individuals with low affinity by (randomly generated) new ones;7. Repeat Steps 2 to 6 until a certain stopping criterion is met.
34 Discrete Immune Network Models (Timmis & Neal, 2001) Initialisation: create an initial network from a sub-section of the antigensAntigenic presentation: for each antigenic pattern, do:2.1 Clonal selection and network interactions: for each network cell,determine its stimulation level (based on antigenic and network interaction)2.2 Metadynamics: eliminate network cells with a low stimulation2.3 Clonal Expansion: select the most stimulated network cells andreproduce them proportionally to their stimulation2.4 Somatic hypermutation: mutate each clone2.5 Network construction: select mutated clones and integrate3. Cycle: Repeat step 2 until termination condition is metInitialise the immune network (P)For each pattern in AgDetermine affinity to each P’Calculate network interactionAllocate resources to the strongest members of PRemove weakest PEndForIf termination condition metexitelseClone and mutate each P (based on probability a)Integrate new mutants into P based on affinityRepeat
35 Somatic Hypermutation Mutation rate in proportion to affinityVery controlled mutation in the natural immune systemTrade-off between the normalized antibody affinity D* and its mutation rate ,
37 Data mining: Problem description More benchmark problem in this caseAssume a set of labelled vectorsClassification
38 AIRS: (Artificial Immune Recognition System) Watkins 2003 Clonal SelectionBased initially on immune networks, though found this did not workResource allocationSomatic hypermutationEventuallyAntibody/antigen binding
39 AIRS: Mapping from IS to AIS Antibody Feature VectorRecognition Combination of feature Ball vector and vector classAntigens Training DataImmune Memory Memory cells—set of mutated ARBs
40 ClassificationStimulation of an ARB is based not only on its affinity to an antigen but also on its class when compared to the class of an antigenAllocation of resources to the ARBs also takes into account the ARBs’ classifications when compared to the class of the antigenMemory cell hyper-mutation and replacement is based primarily on classification and secondarily on affinitySparse in AIS literatureNot as straight forward as initially suspected
41 AIRS Algorithm Data normalization and initialization Memory cell identification and ARB generationCompetition for resources in the development of a candidate memory cellPotential introduction of the candidate memory cell into the set of established memory cells
42 AIRS: Performance Evaluation Fisher’s Iris Data SetPima Indians DiabetesData SetIonosphere Data SetSonar Data SetIris: 3 way classification problem; 150 data items; 5XCV; avg. 3 times; 4 featuresIonosphere: 2-way classification, good & bad radar returns; 34 features; 200 in training, 151 test setDiabetes: 2-way class, has diabetes or not; 10XCV; 8 features; 768 instances totalSonar: 2-way class; 13XCV; 60 features; 16 instances in each test set
43 Classification Accuracy Important to maintain accuracyAIRS1: AccuracyAIRS2: AccuracyIris96.796.0Ionosphere94.995.6Diabetes74.174.2Sonar84.084.9
44 Features No need to know best architecture to get good results Default settings within a few percent of the best it can getUser-adjustable parameters optimize performance for a given problem setGeneralization and data reduction
45 aiNET: Artificial Immune Network for Data Mining
46 Problem description More benchmark problem in this case Assume a set of unlabelled vectorsWe can ask the questions:Is there a large amount of redundancy?Are there any groups or subgroups intrinsic to the data?What is the structural or spatial distribution?
48 Data mining: Immune Network Algorithm 1. Initialization: create an initial random population of network antibodies;2. Antigenic presentation: for each antigenic pattern, do:2.1 Clonal selection and expansion:2.2 Affinity maturation:2.3 Clonal interactions:2.4 Clonal suppression:2.5 Metadynamics:2.6 Network construction:3. Network interactions:4. Network suppression:5. Diversity:6. Cycle: repeat Steps 2 to 4 until a pre-specified number of iterations is reached.Initialization: create an initial random population of network antibodies;2. Antigenic presentation: for each antigenic pattern, do:2.1 Clonal selection and expansion: for each network element, determine its affinity with the antigen presented. Select a number of high affinity elements and reproduce;2.2 Affinity maturation: mutate each clone inversely proportional to affinity. Re-select a number of highest affinity clones and place them into a clonal memory set;2.3 Clonal interactions: determine the network interactions (affinity) among all the elements of the clonal memory set;2.4 Clonal suppression: eliminate those memory clones whose affinity is less than a pre-specified threshold;2.5 Metadynamics: eliminate all memory clones whose affinity with the antigen is less than a pre-defined threshold;2.6 Network construction: incorporate the remaining clones of the clonal memory with all network antibodies;3. Network interactions: determine the affinity (degree of similarity) between each pair of network antibodies;4. Network suppression: eliminate all network antibodies whose affinity is less than a pre-specified threshold;5. Diversity: introduce a number of new randomly generated antibodies into the network;6. Cycle: repeat Steps 2 to 4 until a pre-specified number of iterations is reached.
49 Data mining: Mapping from IS to aiNET Immune SystemaiNETB-cell (antibody)Internal data vectorAntigenTraining data vectorBindingCalculation of Euclidean distanceCell cloningDuplication of internal data vectorsSomatic hypermutationAffinity proportional mutationImmune networkNetwork of internal data vectorsMetadynamicsRemoval and creation of internal data vectors
50 Data mining: Clustering (aiNet) Limited visualisationInterpret via MST or dendrogramCompression rate of 81%Successfully identifies the clustersTraining PatternResult immune network
52 Other Interesting Applications Immune Network for continuous learning (Neal 2002)Track moving data over timeMaintains clusters in absence of patternsUseful for dynamic environmentsContinuous Classificationclassification of interesting/non-interesting sChanging profile of the userMaintain classification accuracyComparable to Naïve Bayes
53 New Trends Danger Theory Could this be useful for Web Mining? Not self/non-self but Danger/Non-DangerImmune response is initiated in the tissues. Danger Zone.This makes it context dependantCould this be useful for Web Mining?
54 Summary Immune metaphors Covered much, but there is much work not covered (so apologies to anyone for missing theirs)Immune metaphorsAntibodies and their interactionsImmune learning and memorySelf/non-selfNegative selectionApplication of immune metaphors
55 The Future Rapidly emerging field Much work is very diverse Framework helps a littleMore formal approach required?Wide possible application domainsWhat is it that makes the immune system unique?More work with immunologistsTheories such as Danger theory, Self-Assertion may have something to say to AIS
56 The Future (2)ARTIST: A Network for Artificial Immune Systems (EPSRC funded network)Work towards:A theoretical foundation for AIS as a new CIExtraction of accurate metaphorsImmune System ModellingApplication of AISTrain PhD studentsFund workshops/meetingsCoordinate and Disseminate UK based AIS research (links to Europe)
57 The Future (hopefully) IT IS: Information Technology Inspired by the Immune SystemFP 6 IP:16 institutions across EuropeCreate a European Library of immune algorithmsTheoretical analysis of AISApplication of AISAutonomous boatImmunoinformaticsWeb MiningModelling of Immune System
58 AIS Resources: BooksArtificial Immune Systems and Their Applications by Dipankar Dasgupta (Editor) Springer Verlag, January 1999.Artificial Immune Systems: A New Computational Intelligence Approach by Leandro N. de Castro, Jonathan Timmis, Springer Verlag, November 2002.Immunocomputing: Principles and Applications by Alexander O. Tarakanov, Victor A. Skormin, Svetlana P. Sokolova, Springer Verlag, April 2003.
59 AIS Related Events in 2003:Special Session on Artificial Immune Systems at the Congress on Evolutionary Computation (CEC), December 8-12, 2003, Canberra, Australia.Special Session on Immunity-Based Systems at Seventh International Conference on Knowledge-Based Intelligent Information & Engineering Systems (KES), September 3-5, 2003, University of Oxford, UK. Second International Conference on Artificial Immune Systems (ICARIS), September 1-3, 2003, Napier University, Edinburgh, UK. Tutorial on Artificial Immune Systems at 1st Multidisciplinary International Conference on Scheduling: Theory and Applications (MISTA), 12 August 2003, The University of Nottingham, UK. Tutorial on Immunological Computation at International Joint Conference on Artificial Intelligence (IJCAI), August 10, 2003, Acapulco, Mexico. Special Track on Artificial Immune Systems at Genetic and Evolutionary Computation Conference (GECCO), Chicago, USA, July 12-16, 2003